Report on methods for demographic projections at multiple levels of aggregation

Skirbekk, V., Prommer, I., K.C., S., Terama, E., & Wilson, C. ORCID: https://orcid.org/0000-0001-8164-3566 (2007). Report on methods for demographic projections at multiple levels of aggregation. PLUREL Report D1.2.1, Module 1: Driving forces and global trends (June 2007)

Preview

Text XO-07-026.pdf Download (301kB) | Preview

Abstract

This report describes methods of internal consistency in population projections at multiple levels of aggregation. The first step in the process is to make the initial assumptions used in the projections at different levels of aggregation consistent. This input-level consistency can be further enhanced by output-level analysis. Comparing the results at the relevant levels of aggregation ensures internal consistency at the output level. Thus, we are able to compare the differences in the age and sex distribution of the population and specific demographic indicators (such as the old age dependency ratio) over various regional levels.

In PLUREL, the national projections will be carried out using the probabilistic method while the regional projections will use deterministic or variant methods. The results at these two levels of aggregation cannot be directly compared one-to-one as there is no simple correspondence between the output variants and the probabilistic range. To avoid any problems arising from this, we develop an index representing the differences in the size and distribution of the population from the variant method to a given percentile in the probabilistic population estimate.

The report discusses various population projection techniques together with their strengths and weaknesses. The relative advantage of specific models for different purposes is discussed forming a selection of models to be used for the population projections in PLUREL: National (NUTS1-0), Regional (NUTS-2) and Case Study Projections (NUTS-5). We conclude that stochastic projections are best suited for national projections, while classic or multiregional cohort-component model projections are likely to be the best choice for the regional projections and for the detailed case study projections.

Actions (login required)

View Item